I'm building an automated investing system from the ground up. The goal isn't a magic money machine — it's the opposite: a disciplined, rules-based engine that removes emotion, protects capital first, and earns trust the slow way — through honest testing, one rung at a time, with a kill-switch always within reach. This page lays out what I'm building, the goals, and exactly how I plan to get there.
This page describes a personal research-and-development project, shared for educational and informational purposes only. It is not investment, tax, or financial advice, not a recommendation to buy or sell any security, and not an offer or solicitation of advisory services. I am not managing money for anyone through this project. Any figures shown are illustrative or drawn from simulations on test data — they are not predictions. Investing involves risk, including loss of principal; past or simulated results do not indicate future performance.
One question drives the whole project:
Can a simple, rules-based system invest a small account honestly — and can I prove it works before risking a single real dollar?
Most "trading systems" are sold on a backtest — a chart of how well they would have done. That proves almost nothing, because it's drawn by someone who already knows how the story ended. This project is built around the opposite belief: a backtest's only honest job is to kill bad ideas cheaply. Real proof comes forward, in real time — so that's what it's engineered to do.
Six commitments shape every design decision. They're the reason the system is trustworthy by construction, not by promise.
Protect capital first.
A built-in crash filter moves the account to safe assets (bonds, T-bills) when markets are falling. Losing less in bad years matters more than winning big in good ones.
Prove before you trust.
Real money is the last step, never the first. The system climbs a ladder of escalating tests — and most ideas are supposed to die early. That's the point.
No emotion, no guessing.
Every decision is a fixed rule, written in code. That makes the system repeatable, auditable, and immune to fear and greed — the two things that wreck most investors.
Cheap and tax-smart.
Low turnover, large liquid funds, and a tax-advantaged account keep fees and taxes near zero. As a CPA, I know the tax drag quietly eats more returns than most people realize.
Transparent and reversible.
The system reads market data but cannot place a trade until a human deliberately turns that on. A kill-switch to halt everything is a permanent fixture, not an afterthought.
What's tested is what trades.
The exact same decision-making code runs in testing and in real life. No bait-and-switch between the version that looks good and the version that actually trades.
The system has three parts that work together:
A read-only assistant that reads company filings and market news and drafts plain-English summaries — so the research is fast, consistent, and never skipped. It can read, but it cannot trade.
A fixed set of rules that, once a month, ranks a small handful of well-known funds by their recent strength and holds the leaders — but only if they're also beating the safety of cash. If nothing is, it rotates entirely to safety and waits out the storm.
A simulator that replays history and live-but-pretend markets, charging realistic costs, to test the rules honestly before any real money is involved.
The key idea that keeps it honest: one set of rules, tested three ways. The very same code is run on history, then on play money, then — only if it earns it — on real money.
flowchart LR
BRAIN["🧠 ONE set of rules
the strategy"]:::brain
BRAIN --> H["Runs on
HISTORY
backtest"]:::h
H -->|"must pass first"| P["Runs on
PLAY MONEY
paper trading"]:::p
P -->|"must pass first"| L["Trusted with
REAL MONEY
live · gated"]:::l
classDef brain fill:#1b2330,stroke:#56d4dd,stroke-width:1.5px,color:#e6edf3;
classDef h fill:#13202e,stroke:#58a6ff,color:#e6edf3;
classDef p fill:#0f2424,stroke:#56d4dd,color:#e6edf3;
classDef l fill:#0f2417,stroke:#3fb950,color:#cfe9d6;
This is the heart of the plan. Each step is harder to fool than the last, and the strategy must pass each one before it's allowed to climb to the next. Real money sits at the very top — gated behind everything below it.
The system connects to the brokerage and reads the account and live market data — and by design cannot place a trade. Safety is the floor, not a feature added later.
The rules are written in code, and the backtester runs end-to-end. On a stress-test scenario with a built-in market crash, it does what it's designed to do: it rotates to safety and rides out the downturn.
Wiring up the plain-English research briefs, and moving the backtest from simulated data onto real market history — then the walk-forward test on data the rules have never seen.
Run the strategy live against real markets with pretend money, on a monthly schedule, and compare what it actually does to what the simulation predicted.
Only if paper trading holds up. Real money starts deliberately small, behind a manual switch, with hard loss limits and an automatic halt. Scale only on evidence.
The most useful thing I can say about a small starting account is the least exciting: in the early years, how much you contribute matters far more than how clever the strategy is.
That's the whole philosophy in one line: get the boring things right — risk control, costs, taxes, consistency — and let time do the rest. A strategy is a hypothesis to be tested, never a promise.
For the curious, two companion pages get into the engineering and the honest-testing methodology: